This paper provides an international perspective on results for students on measures of 21st century skills in the International Study of City Youth (ISCY). The paper uses multi-level modelling techniques to explore within and between school differences in 21st Century skills across participating cities. This is supported by detailed case studies of individual cities provided in the other papers that are part of the symposium. The particular significance of this paper lies in its focus on 21st century skills. International comparison to date has focused primarily on cognitive skills, which are more easily assessed across international contexts. However, international comparisons “have not yet been able to evaluate… competencies which are of increasing importance, such as the capacity of students to relate well to others, to manage and resolve conflicts or to respect and appreciate different values, beliefs or cultures” (Schleicher 2010, p. 493). This paper answers the need for international comparisons of non-cognitive skills, given their importance to learning and post-school outcomes. The paper explores how much variation exists in measures of 21st century skills across participating cities, and between schools within each system. It then sets out to identify the school-level practices, as well as differences in curriculum and program structures, and features of school organisation, that shape differences in levels of 21st Century skills (skills such as conscientiousness, academic application, hope, perseverance, and ICT literacy). The source of data is the ISCY baseline survey conducted in 12 cities around the world in 2013-2014. Data from the five cities in the symposium involves over 13,000 students across 137 schools. Multi-level modelling is used to measure student, school and city differences. There is widespread recognition of the value of international comparisons of education systems. Such comparisons overcome the impossibility of undertaking “experimental” research at a system level, by enabling different systems to illuminate the impact of different educational structures and programs on student outcomes (Schleicher 1995, p. 217). At the same time, such studies require a careful balance between the similarities that facilitate comparison, and the differences that produce errors in comparative studies (ibid). Multi-level analysis is used to help isolate the factors that have a universal impact on student outcomes, regardless of city or school context.

To be provided